Slashdot Mirror

The following was written by Slashdot Reader Matthew Priestley, who, despite his email address, is a pretty cool guy Honest News in the Slashdot Decade
In this paper, we discuss the nature of biased and unbiased news in terms of 'trust decisions', using the cryptographic sense of that phrase. We examine the biases in modern media and identify their causes. Two examples of community news services are examined: Slashdot.org, and FreeRepublic.com. (0) From this analysis we derive a model of community news.

Disclaimer: The author of this paper works for Microsoft, but his opinions may not be the opinions of Microsoft. In fact, they aren't. The author hereby declares that nobody important at Microsoft is even aware of his existence, and that he is about as significant to Bill Gates as a single bacterium in your colon is significant to the weather in France.

0 Introduction
There is a malaise of distrust among news consumers. In recent years the number of news outlets has dwindled due to mergers and attrition, leaving information consumers with a scrawny range of choice. As the global quantity of information grows at a jaw-dropping rate, individuals increasingly despair of their ability to filter the news without aid from massive corporations.

Almost half of adults have little or no trust in media agencies (1), yet almost all delegate news collection to companies they will condemn if asked. When consumers knowingly act against their own interests, a form of coercion must be in operation. In the case of news, this coercion is a stranglehold enjoyed by media companies over filtered information. If their services are not accepted, the consumer sinks in a sea of data. In a world in which no one can process all the news and still enjoy a full life, having all information is as useless as having no information at all.

1 Nature and weakness of trust decisions
The selection of a news-filtering agency resembles what is called in cryptology a 'trust decision'. Briefly, a trust decision is a choice made by the user to validate another user's digital certificate. By assigning trust to the certificate, any content signed by that certificate becomes, in a limited sense, trustworthy. (2)

It is burdensome to evaluate the trustworthiness of every certificate, and a typical user lacks the expertise to investigate each exhaustively. For this reason, most users choose to trust a Certification Authority or CA, a central agency empowered to make trust decisions on their behalf. By endowing a single node with the power to filter certificates, the user is spared this chore. (3)

This process is analogous to the decision to accept news from an established information outlet. It would require an unreasonable effort and scads of time for any individual to audit all the news. Apart from sheer volume, appraising facts often requires background familiarity. Sources must be checked, viewpoints solicited, and impact considered. It becomes clear that this is no task for a person who hopes to conduct, for example, a life on the side. Hence the necessity of the trust decision.

Due to the exhausting claims of evaluating news, authority to filter information must be delegated.

2 Sources of bias in modern media
2.1 Opinion pollution
That trust decisions are subject to predation should be apparent. The most evident form of bias is opinion pollution, in which the subjective feelings of a reporter taint the news. Such bias may either systemic, or it may be the fault of "rogue" reporters, or both.

This form of bias is trivial to establish. In a July 8th article discussing a verdict against tobacco companies, the New York Times dwells on the volume of damning evidence presented by the plaintiffs. The deformities of the smokers are described, and the article drops a helpful tip about joining the suit. (4) Covering precisely the same event, the Wall Street Journal scrupulously avoids discussing the smokers, save to describe their organizers as 'flamboyant'. The spectre of a flooded court system and billions in costs is raised multiple times, and the guilty verdict categorized as a legal 'aberration'. (5)

This form of trust violation can be characterized in two ways. If the tolerance for personal beliefs in the news is not widespread, but isolated to a few reporters, then officials of the corporation have delegated their authority unwisely. An organization that is otherwise trustworthy will eventually correct this error. If the corruption runs throughout, however, then the consumer's initial trust decision was poor. In either event, ongoing opinion pollution can only be sustained by broad organization-wide consensus on the value of certain ideas.

Opinion pollution is a trait of homogeneous groups.

2.2 Advertising revenue and corporate ownership
Often overlooked as a source of bias is the murky relationship between news providers and advertisers. The age-old subscription model has fallen by the wayside, unable to compete with advertiser-funded services that appear to offer information for free. (6)

One fallacy is that advertising flows toward high readership, rewarding popularity with success. In reality, corporations are not interested in buyers, not readers. The Daily Herald, a worker's paper in 1960's England, boasted a readership of 4.7 million the year of its demise - nearly double that of the Times, the Financial Times, and the Guardian combined. (7) But the Herald's readers were demi-socialists, and failed to support the very businesses keeping their paper alive. The advertising money melted away.

A look at subscription income and advertising income emphasizes the dwindling importance of readers. A copy of The Washington Post costs as little as 24 cents a day. By contrast, one inch of black-and-white advertisement in the paper commands $257.55. (8) Economically, it would be more prudent for the Post to alienate 1000 readers than one business buying a daily inch of print. If the lost readership were confined to non-buyers, advertising rates would not even have to drop. When profit per advertiser squashes profit per consumer, the business of advertiser-funded information outlets becomes not the sale of information, but the sale of a receptive audience.

The situation is aggravated when a large corporation owns the news-filtering outlet. Most fans of TV news are unaware ABC is owned by Disney, NBC by GE with investment from Microsoft, and CBS by Westinghouse Electric. Stories critical to these interests are treated gingerly in the news. (9)

Reliance on advertising or corporate ownership selects for news that is business-friendly. High readership is no exemption.

2.3 Feeder authority
Any reader who has attempted to wrest information from the government is aware of its inertia. Similarly, the PR departments of businesses are known for their unhelpful volubility. In the first case the problem is information deficit, in the second it is disinformation glut, but ultimately the predicament is the same.

The situation is no different in a modern newsroom. Effective reporters are those who have established personal relationships with 'sources' inside various institutions who feed them privileged information. These reporters are superior information gatherers because they may ask questions that typically are rebuffed.

Without the goodwill of their 'feeders', even competent journalists drown in a sea of flack. Should an information gatherer alienate an important feeder, the gatherer is instantly severed from a pool of developing information. Pains are taken to ensure feeders are pleased with the treatment of their comments in published accounts. (10) This creates an unhealthy environment for the analysis of news. If an information outlet were to criticize the statements of a feeder, or if fallacies or lies were exposed in the feeder's reasoning, the potential effect on the outlet would be calamitous. This allows the feeder to make use of information outlets as occasional distributors of propaganda, knowing that refusal is unlikely.

Information from a small number of feeders may be propagandized.

3 News distribution over the Internet
Slashdot.org and FreeRepublic.com are representatives of a new class of news filter. While using these sites, consumers alter the fundamental structure of their trust decision. Rather than inhabiting a descending tree, in which trust is derived from progressively higher and fewer nodes, a Slashdotter or Freeper distributes their trust. In a distributed trust model, each consumer inhabits a single node in a formless but highly connected graph. Central authority is weak, participants are anonymous, and all nodes perform small amounts of voluntary labor.

3.1 Slashdot.org
Recently thrown mainstream as a gathering spot for Linux advocates, Slashdot.org has a large and devoted following of geeks and technophiles. Interestingly, because of its adherence to transparency and peer review, Slashdot has evolved a news system that defeats several of the biases described above. Slashdot is the conceptual descendent of the Internet newsgroup and the old-timer's BBS. Members log in to the web board and select one or more current items to discuss, then post their reactions.

3.1.1 Successes of the Slashdot model Participants on Slashdot are only identifiable if they wish to be. Widespread use of aliases insulates participants from real-world reprisal - a Slashdotter may criticize the government, their employer, or other feeders with small risk. Handle-use also renders a state of meritocracy on Slashdot. Comments and topic submissions are judged by their own merits, since little is known about their real-world source. Aliases grow trusted in the forum as a result of their owner's contributions. Deprecated aliases have only themselves to blame.

Members submit topics on Slashdot, and those with promise are posted to the forum. By distributing the labor of reporting, the process of information collection becomes inexpensive, and the likelihood of discovering important news increases - much like the 'Have you seen this child?' ads on milk cartons. (11) When the system requests voluntary labor, it is limited to tasks costing only a few mouse clicks. The decision of what is 'newsworthy' is also simplified, since an audience member has provided the item. If each registered Slashdot member contributed only 1 minute per day, their efforts would sum to 1083 work-hours of labor - absolutely free.

Relinquishing trust to anonymous lurkers appears foolhardy, but as randomness grows, so does quality. The web demographic is a straw poll in the worst sense of the term (12), but there are tide pools of demographic validity if groups are narrowly defined. When a site achieves a certain level of notoriety, Slashdot for example, a cross-section of users may fairly be said to represent its supporting community, in this case idealistic geeks. An information consumer is not interested in topics useful to the average person; rather they are interested in what is useful to people like themselves.

No opinion is authoritative until it runs the Slashdot gauntlet. Members comment on topics, share experiences, and take potshots at sloppy reasoning. This is more egalitarian than the feedback model of magazines, TV, or books. In those cases, if a retort is even possible, it is run in the following issue, with no guarantee to reach the original audience. On Slashdot, user comments frequently upstage the 'official' news, and it is a testament to their quality that reading the primary source is often unnecessary. Because most topics excite a gamut of opinions, Slashdot defeats the threat of opinion pollution.

To tame dull or off-topic comments, Slashdot members are randomly empowered to moderate the 'score' of remarks. Moderators are chosen by the system with a preference towards regular but not ubiquitous readers. Comments that gain the approbation of everyday participants gradually move up through statistical effects. Pointless comments sink into oblivion. Visitors to the forum may choose their own threshold of dependence on this ratings system. On Slashdot, the uniform opinions of classic information outlets are rare.

Finally, the scripts and HTML that run Slashdot are released to the community. This ensures, within reason, that the site truly operates as billed, as well as opening the code to all the benefits of open source.

3.1.2 Failings of the Slashdot model
Among its positive effects, anonymity damages credibility. If Secretary of State Madaleine Albright posted a remark on technology export limitations, her opinion would be more significant than had 'DrDeath' typed precisely the same opinion. Validation of real-world credentials can be desirable. One solution would be to support either the S/MIME or PGP signing standards as a user option. A hash of important messages could be included with the post, thereby validating the identity of the signer. (13)

No Slashdot participant receives a handle until they submit an e-mail address to the Slashdot central authority. Those who do not may participate as 'Anonymous Cowards'. AC's suffer numerous disadvantages, not the least that their posts begin at a lower score. Though this distinction discourages meddling from non-regulars, it is risky. Regular members are no less anonymous or even cowardly than AC's, save that they have disclosed their private information to the Slashdot central authority. This makes criticism of the authority more difficult, since critical remarks are safe only as an AC post from a lab computer, which is immediately scored down.

There is one departure on Slashdot from democracy. While consumers do submit the discussion topics, these are dropped into an administrative black box, unseen until a few emerge handpicked by the central authority. Inside the 'box', a small number of humans, vulnerable to self-interest, choose which of the topics will be news. In theory, the authority could even replace submitted topics with its own. A better system would be an open one, moderated in the same manner as user remarks. Along with their ration of remark-points, moderators would be given a supply of topic-points, which could be spent on proposed topics in a pool. Users could set topic thresholds in the same manner that they set thresholds for remarks. This method would be self-policing and eliminate tedious work for the central authority. (Update: 07/16 01:15 by CT : See the Slashdot FAQ for the reason that I've decided not to do this)

Slashdot is funded by banner advertisements, and on 6/29/99 announced that it had been acquired by Andover.net. (14) While there is little danger of the various Linux distros exerting pressure as yet on Slashdot, and while Andover rarely appeared on Slashdot in the past, nonetheless these developments cast a shadow on the impartiality of the community forum. Is it less likely that a story criticizing Sony will be run when an advertisement for the Sony AIBO adorns the top banner? What would become of stories damaging to Andover? Members should be alert for signs of conflicting interest.

3.2 FreeRepublic.com
Similarly evolved, although less highly automated, is FreeRepublic.com, a forum for the exchange of conservative commentary. FreeRepublic is similar to Slashdot in appearance and general design. We will focus on their differences.

3.2.1 Successes of the FreeRepublic model
FreeRepublic's most notable trait is the freedom members enjoy in topic selection. Power is so far in their hands that every member may post any topic they choose, resulting in dozens of discussed topics per day. A true distributed trust network has no single point of entry. Since the number of daily articles is finite, any given node in a sea of nodes has negligible influence. Individuals may be bought or coerced, but since the merits of each contribution are peer-reviewed and peer-diluted, successful corruption must be hugely widespread. The resources needed to influence a majority of users would be prohibitive, and only dubiously worthwhile. Once accomplished, the forum would cease to serve the needs of valid members and would naturally dissolve. Attempts to corrupt distributed news forums are by nature self-defeating.

FreeRepublic reaps no funding from advertisement or corporate ownership. The site is fed by out-of-pocket donations from participants. Though it should be noted that FreeRepublic's supporting community stereotypically has more disposable income than the average netizen, even so the site is accountable to none save its members. When the object of a news outlet is the aggregation of money, it should be unremarkable when money supersedes the pursuit of information. But in a community forum, participants have no aim other than valuable and convenient news.

Participants on FreeRepublic meet physically, organize in chapters, and crusade in the real world to accomplish their aims. There is little risk to anonymity, since there is no need to divulge onscreen handles. Provided chapters are small and independent, the inevitable discussion of principles will not even dampen diversity of opinion, which could expose the forum to opinion pollution. Participants also leave the meetings with a sense of community, which increases their voluntary labor.

3.2.2 Failings of the FreeRepublic model
Although a blessing, complete freedom of topic selection is also a curse. At times of peak activity, two successive clicks on Refresh may result in two completely different topic lists. Crackpots frequently post and their topics slide off the page untouched by regulars. There is much duplication as news breaks. Most topics receive fewer than twenty comments, reducing the effects of peer-dilution and peer-review. All these problems could be resolved if FreeRepublic were to transition to the scoring-based topic selection approach recommended previously.

FreeRepublic has no moderation method for comments, and consequently all remarks carry equal weight. In its absence, opinions win by volume or position near the top of the remark list rather than insight or appeal to the median qualities of the community. Corruption of an unmoderated forum is trivial given fifty aliases and sufficient time.

On FreeRepublic, community participants are not permitted to comment or post discussion topics unless they are logged on. This is an extreme case of Slashdot's Anonymous Coward dilemma. No contribution can be made to the forum without being noted by the FreeRepublic central authority. There is no guarantee the central authority will not terminate or diminish the accounts of those who criticize its practices.

Finally, FreeRepublic is closed source. Though the site is more static than Slashdot, what scripts it has are not disclosed to the forum. Members must take it on trust that no back doors lurk in the code.

4 Issues in Internet news distribution
4.1 The trouble with enthusiasm
One trait of both Slashdot and FreeRepublic is that their populations contain a percentage of zealots. This fact attracts the attention of non-members and ensures the continued participation of long-standing ones. While allegiance to a specific viewpoint is in no way an exclusionary criterion on Slashdot or FreeRepublic, most users share a common opinion on a few controversial issues. This may reflect the fact that contentious topics generate the most passionate interest.

Regrettably, this bond introduces a capacity for bias. Most information processed on a trust graph will lie outside the emotional boundaries, allowing peer-review and peer-dilution to ensure honest news analysis. But when discussion touches on a 'hot button' topic, rampant uniformity of opinion eliminates these safeguards.

FreeRepublic may safely be termed incapable of objective thought when the topic of President Clinton is broached. One recent post discussing Clinton's attendance at the World Cup bore the helpful keywords 'CLINTON RAPIST EVIL SLEAZY TRAITOR'. (15) Similarly, the high quality of discourse on Slashdot disintegrates when Microsoft enters the headlines. Both communities may be absolutely correct in their opinions on these topics, but the mere fact of consensus mimics the effects of corruption and degrades the community information filter. Whether it is desirable or even possible to generate a community forum without this sort of bias is a question for further debate.

4.2 Overcoming feeder bias
Although incisive analysis may overcome the flaws in a poorly written news article, community forums are ultimately limited by their feeders. These feeders are not usually primary sources, except in cases where significant documents are available online. Far more common is the linking of news articles from established information filtering corporations. The question arises whether community news efforts can surmount partiality on the part of the original reporters.

The answer appears to be yes. When CPU-maker AMD recently released comparisons between its chips and those of rival Intel, Slashdot was quick to dissect the biases in presentation and supply the necessary omitted background. (16) However, it should be noted that processors are a topic enjoying high familiarity among the technical elite who visit the site. Had the discussion been on the political condition of Nicaragua, results would be sketchy at best. Fortunately, community information forums are inherently unlikely to encounter this dilemma. Since the group as a whole selects topics, discussions lying outside the expertise of the majority are rare. A more difficult question is this: will community news replace traditional news outlets, or merely supplement them?

5 Conclusion
Community information filters are a novel approach to news. Trading on the principles of self-interest and distributed trust, they levy the expertise of thousands into producing honest, cheap daily news. In a world where command of information is rapidly becoming the root of institutional power, distributed trust graphs refocus information upon the needs of the citizen. While they remain in a state of infancy, the rise of sites such as Slashdot and FreeRepublic herald the demise of traditional information flows. We have entered the Slashdot decade, and only time will judge our success.

6 References
(0) http://www.slashdot.org, http://www.freerepublic.com
(1) http://www.gallup.com/poll/releases/pr990108.asp
(2) http://www.rsa.com/rsalabs/faq/html/4-1-3-11.html
(3) E.g. http://www.thawte.com
(4) "Tobacco Industry Loses First Phase of Broad Lawsuit", New York Times, 6/8/99
(5) "A 'Class' Trial Finds Tobacco Firms Liable; Big Payments May Follow", Wall Street Journal, 6/8/99
(6) Cable is an exception. The means of distribution in cable are monopoly-owned, preserving cable from direct competition with TV.
(7) Herman & Chomsky, Manufacturing Consent, Pantheon Books, p15, [cf.]
(8) As of July 1999, Washington Post, http://www.washingtonpost.com/wp-srv/guide/sub/sub.htm, http://adsite.washpost.com/rates/retail/fullrun.html
(9) http://www.fair.org/media-woes/media-woes.html
(10) E.g. http://independent.org/tii/content/events/f_macarth.html
(11) http://www.missingkids.org
(12) http://www.ntia.doc.gov/ntiahome/digitaldivide
(13) http://www.rsa.com/rsalabs/faq/html/2-2-2.html
(14) "Slashdot Acquired by Andover.Net"
(15) "Clinton hopes for soccer diplomacy"
(16) "Athlon Benchmarks Out"